NEW RESEARCH

Treatment of Children With Attention-Deficit/Hyperactivity

Disorder (ADHD) and Irritability: Results From the
Multimodal Treatment Study of Children With ADHD (MTA)
 ndez de la Cruz, PhD, Emily Simonoff, MD, James J. McGough, MD,
Lorena Ferna
Jeffrey M. Halperin, PhD, L. Eugene Arnold, MD, MEd, Argyris Stringaris, MD, PhD, MRCPsych

Objective: Clinically impairing irritability affects 25% to and behavioral treatment was superior to community care
45% of children with attention-deficit/hyperactivity dis- and to behavioral treatment alone, but not to medication
order (ADHD); yet, we know little about what inter- alone. Irritability did not moderate the impact of treat-
ventions are effective in treating children with ADHD and ment on parent- and teacher-reported ADHD symptoms
co-occurring irritability. We used data from the Multi- in any of the 4 treatment groups.
modal Treatment Study of Children With ADHD (MTA)
to address 3 aims: to establish whether irritability in Conclusion: Treatments targeting ADHD symptoms are
children with ADHD can be distinguished from other helpful for improving irritability in children with ADHD.
symptoms of oppositional defiant disorder (ODD); to Moreover, irritability does not appear to influence the
examine whether ADHD treatment is effective in treating response to treatment of ADHD.
irritability; and to examine how irritability influences
Clinical trial registration information—Multimodal
ADHD treatment outcomes.
Treatment Study of Children With Attention Deficit
Method: Secondary analyses of data from the MTA in- and Hyperactivity Disorder (MTA); http://www.
cluded multivariate analyses, and intent-to-treat random- clinicaltrials.gov; NCT00000388.
effects regression models were used.
Key Words: irritability, attention-deficit/hyperactivity
Results: Irritability was separable from other ODD disorder, oppositional defiant disorder, treatment out-
symptoms. For treating irritability, systematic stimulant comes
treatment was superior to behavioral management but not
to routine community care; a combination of stimulants J Am Acad Child Adolesc Psychiatry 2015;54(1):62–70.

C
linically impairing irritability affects 25% to 45% of general population rates.8 This raises the question of how
children with attention-deficit/hyperactivity disor- best to treat the subgroup of children with ADHD and
der (ADHD)1; yet, the evidence base for treatment irritability.
selection in the presence of irritability remains thin. This One approach to this question is to consider irritability as
article addresses this knowledge gap by analyzing data from one of the manifestations of behavior problems that are
the Multimodal Treatment Study of Children With ADHD typical of ADHD. Indeed, irritability is characteristic of
(MTA), a large randomized trial comparing various treat- children with oppositional defiant disorder (ODD), which is
ment modalities among children with ADHD.2 highly comorbid with ADHD.9 Substantial evidence sug-
ADHD is among the most common child psychiatric gests that stimulant treatment also reduces ODD symptoms
disorders worldwide.3 It is defined by chronic, pervasive, in those with ADHD2,10,11 and that parenting interventions
and impairing symptoms of inattention and hyperactivity/ may also be useful.12 However, this is an assumption that
impulsivity.4 Although irritability, defined by temper out- needs to be explicitly tested, as mounting evidence indicates
bursts and proneness to anger,5 is not a diagnostic criterion that irritable mood is distinct from other, typically head-
for ADHD, it is a common presentation in this clinical strong, behaviors characteristic of ODD. This distinction is
group6,7 and is listed under the associated features of ADHD reflected in the DSM-54 and is based on research showing
in the DSM.4 In an epidemiological study, 38% of children that irritability is separable from headstrong behaviors (e.g.,
with ADHD had irritable mood, nearly 10-fold higher over argumentativeness, noncompliance, and rule breaking) by
virtue of its multivariate structure,13 longitudinal course,14,15
external predictions,16 and genetic associations.17 In partic-
ular, irritability predicts subsequent depression and gener-
alized anxiety, whereas headstrong behaviors predict
Clinical guidance is available at the end of this article.
subsequent delinquent behaviors.16 If irritability is clinically
and etiologically distinct from other behavior problems, ir-
Supplemental material cited in this article is available online. ritability in ADHD may also require distinct treatment
compared to headstrong behaviors. However, there is little

JOURNAL OF THE AMERICAN ACADEMY OF C HILD & ADOLESCENT PSYCHIATRY

62 www.jaacap.org VOLUME 54 NUMBER 1 JANUARY 2015
 TREATMENT OF ADHD AND IRRITABILITY

research on the distinctions between these 2 groups of Behavior Checklist (CBCL) dysregulation profile, suffered
symptoms in children with ADHD13 or evidence about how more morbidity at study onset, yet they also responded to
best to treat irritable children who have ADHD. standard ADHD treatment without suffering more side ef-
Here we use data from the MTA to address these ques- fects compared to children without manic symptoms.25 This
tions by examining 3 aims. First, we wanted to establish the suggests that standard treatment may benefit severely
robustness and clinical relevance of irritability in the MTA. affected children with ADHD; however, the inclusion of
In particular, it is important to know whether irritability in anxiety/depression, aggression, and attention in the defini-
children with ADHD can be distinguished from other typical tion of manic symptoms makes it difficult to estimate the
symptoms of oppositionality, namely, headstrong symptoms. effects of treatment on irritability and on how the presence of
We hypothesized the following: that irritability would be irritability may moderate treatment response in children
separable from headstrong behaviors in multivariate analyses; with ADHD. This is particularly important given the sepa-
that irritability would have different external correlates/ rability of irritability from other behavior problems. Hence,
consequences (building upon prior investigations,14 we ex- in this study, we tested whether the response to MTA
pected that irritability would significantly differentially treatments, including medication, behavioral treatment, and
predict internalizing symptoms and disorders such as the combination, varied according to the level of irritability.
depression and anxiety, whereas the headstrong dimension Our expectation was that while irritable children would
would differentially predict conduct problems, such as show higher levels of ADHD symptoms, they would
conduct disorder [CD]); that irritability would show suffi- respond similarly to children low on irritability. In partic-
cient longitudinal continuity in children with ADHD such ular, we expect that the previously demonstrated superiority
that it could be differentiated from headstrong symptoms; of the medication management over the community com-
and that irritable and headstrong behaviors each would parison and the behavioral treatment arms in the MTA2
contribute independently to impairment. would remain even when accounting for levels of irritability.
Our second aim was to ascertain how irritability in
ADHD responds to treatment with stimulants and/or
behavioral therapy. Clinical experience and prior results METHOD
from randomized controlled treatment studies in chil- Participants and Procedure
dren18-20 suggest that stimulant treatment may be useful to A total of 579 children meeting diagnostic criteria for ADHD
treat irritability in ADHD and should be considered as a Combined Type were recruited from 6 different US sites and
first-line treatment1; however, the evidence is somewhat randomly assigned to 1 of the following 4 groups: medication
mixed. Two randomized controlled trials comparing management (“MedMgt”; n ¼ 144; 24.9%); behavioral treatment
(“Beh”; n ¼ 144; 24.9%); combined treatment (“Comb”; n ¼ 145;
amphetamine and placebo found no beneficial effect of the
25.0%); or treatment-as-usual community comparison (“CC”; n ¼
medication on a broad range of emotional problems, and 146; 25.2%). The first 3 groups were treated for 14 months in spec-
some studies have found that amphetamine preparations ified protocols. Briefly, MedMgt consisted of a 1-month double-blind
increase irritability and lability.21 On the other hand, it has titration with methylphenidate for best dose, progressing to an open
recently been shown that children with ADHD and behavior titration with other drugs, such as d-amphetamine, pemoline, or
problems that are closely linked to irritability probably imipramine if methylphenidate was unsatisfactory. Beh consisted of
respond to behavioral treatment.22,23 Similarly, a recent intense, multi-component individual and group parent training;
meta-analysis of randomized controlled trials12 has provided teacher consultation; a child-directed, 8-week, full-time summer
evidence from blinded outcomes that behavioral in- treatment program; and use of a 12-week, half-time classroom
terventions improve parenting and reduce childhood behavioral specialist. Comb integrated the MedMgt and Beh stra-
tegies, with more extensive assistance from the behavioral therapist
conduct problems in ADHD. To address this matter, we
examine 2 competing hypotheses: 1 hypothesis based on the
preliminary findings above that suggest stimulant treatment FIGURE 1 Parent-reported irritability response to multimodal
could be helpful; and the other hypothesis, which is that treatment in the 4 treatment groups. Note: Beh ¼ Behavioral
because irritability can be a component of behavior problems treatment; CC ¼ Community Comparison; Comb ¼ Combined
such as ODD, it may respond well to behavioral treatment. treatment; MedMgt ¼ Medication management.
The third aim was to establish whether the response to
treatment of children with ADHD and irritability differed
from that of children without irritability. There is surpris-
ingly little research in this area, although a previous study
using MTA data indicated that symptoms of mania do not
influence the treatment response to methylphenidate or its
side effect profile in children with ADHD.24 However, this
investigation used the 1-month methylphenidate titration
trial subset of the MTA and therefore did not include the
comparisons with the behavioral interventions nor the
treatment outcome after 14 months. A subsequent study by
Galanter and colleagues, also using MTA data, showed that
children with manic symptoms, as defined using the Child

JOURNAL OF THE AMERICAN ACADEMY OF C HILD & ADOLESCENT PSYCHIATRY

VOLUME 54 NUMBER 1 JANUARY 2015 www.jaacap.org 63
 
FERNANDEZ DE LA CRUZ et al.

to assist in medication adjustment and information from the phar- measure was based on parent-report, given that parents, as
macotherapist to aid in decision making about escalation of compared to teachers, are rated as more useful informants of chil-
behavioral interventions. The fourth group (CC) was referred for dren’s emotional problems.31,32
community treatment of the parents’ choosing. The mean age of the A headstrong dimension was generated by adding up 4 ODD
children at baseline was 8.5 years (SD ¼ 0.8 years, range ¼ 7–10 items on the parent-reported SNAP: “Argues with adults,”
years), and 114 children were female (19.7%). Ethnic composition of “Actively defies or refuses adult requests or rules,” “Does things
the sample included 60.8% of white ethnicity, 19.9% African deliberately that annoy other people,” and “Blames others for his or
American, and 19.3% Hispanic, racially mixed, or from other ethnic her mistakes or misbehavior.” The scores ranged from 0 to 12. The
origins. Treatment groups did not differ significantly at baseline on only ODD item not used for either the irritability or the headstrong
gender, ethnicity, IQ, comorbidity, Conners Parent and Teacher scales was “Is spiteful or vindictive.”
Rating Scales scores, impairment, and medication for ADHD before The CBCL was used to assess general psychopathology. The
the study. The only significant difference was age, although all CBCL is a parent-report checklist mapping onto multiple aspects of
participants were actually in a tight age range (the youngest were in psychopathology over a 6-month period.33 Children’s global
the behavioral treatment group: mean age ¼ 8.3 years; and the impairment was measured by using the Columbia Impairment
oldest were in the medication management group: mean age ¼ 8.6 Scale–Parent Version (CISP) questionnaire.34
years). At least 1 parent and all participants in the original MTA
study provided written permission and assent for participation
before initiation of any study procedures as approved by each site’s Data Analysis
institutional review board. For the present study, all of the partici- Data analyses for each specific aim were as follows: 1) To establish
pants from the original study were included. Additional details whether irritability was independent from other ODD symptoms in
about the sampling and the procedures in the MTA have been the MTA sample, we proceeded in 4 ways. First, we explored dif-
widely described elsewhere.2,26,27 ferences in the multivariate structure using a confirmatory factor
analysis (CFA) comparing 2 models (1 versus 2 factors, namely ir-
ritability and headstrong behaviors). Second, we explored the lon-
Measures gitudinal continuity of each measure (irritability and headstrong
The measures relevant to our study are described below. A behaviors) using path analysis. Third, we explored whether irrita-
comprehensive description of the assessment measures used in the bility and headstrong behaviors had different correlates in linear
MTA has been described elsewhere.2,28 regression models in which the 2 variables were introduced as
ADHD symptom severity was measured by using the mean predictors, Finally, we ran a linear regression in which irritability
score of items 1 to 18 of the parent- and the teacher-reported and headstrong behaviors were predictors of impairment to test
Swanson, Nolan, and Pelham (SNAP) rating scale,29 which includes whether the 2 dimensions contributed independently to it. This is
the inattention and hyperactivity/impulsivity subscales. important because impairment can be independent of symptom
In accordance with previous studies,14,30 a measure of irritability severity.35 Significant differences between estimates were judged
was generated by adding up the following 3 ODD items on the based on nonoverlapping 95% CI. 2) To test the hypothesis that
parent-reported SNAP: “Loses temper”; “Is touchy or easily MedMgt would be superior to Beh in treating irritability, we ran an
annoyed by others”; and “Is angry and resentful.” The scores ranged intent-to-treat (ITT) random-effects regression analysis similar to the
from 0 to 9. In addition, a categorical irritability outcome was original primary analyses but with irritability as the outcome, and
generated using a median split into high and low irritability and was time (including baseline, 3-month, 9-month, and 14-month assess-
used for purposes of illustration in several figures. The irritability ments) and treatment group as predictors, as well as the interaction

TABLE 1 Irritability Response to Attention-Deficit/Hyperactivity Disorder (ADHD) Treatment in the 4 Treatment Groups
Irritability Scores (SNAP Parent Report)

3-Month 9-Month 14-Month

Baseline Assessment Assessment Assessment
Within-Group Effect Size
Treatment Group Mean SD Mean SD Mean SD Mean SD (Baseline to 14-Month Assessment)
Community comparison 4.40 2.43 3.77 2.53 3.83 2.56 3.28 2.18 0.48
Medication management 4.40 2.65 2.81 2.37 2.75 2.26 2.80 2.39 0.63
Behavioral treatment 4.11 2.44 3.39 2.31 3.25 2.20 3.11 2.41 0.42
Combined treatment 4.26 2.41 3.02 2.05 2.64 2.08 2.35 2.22 0.82

Between-Group Effect Size

Treatment Group Comparison (Baseline to 14-Month Assessment)
Medication management vs. Community comparison 0.20
Behavioral treatment vs. Community comparison 0.05
Combined treatment vs. Community comparison 0.34
Medication management vs. Behavioral treatment .24
Combined treatment vs. Medication management .13
Combined treatment vs. Behavioral treatment 0.38
Note: SNAP ¼ Swanson, Nolan, and Pelham rating scale.

JOURNAL OF THE AMERICAN ACADEMY OF C HILD & ADOLESCENT PSYCHIATRY

64 www.jaacap.org VOLUME 54 NUMBER 1 JANUARY 2015
 TREATMENT OF ADHD AND IRRITABILITY

timetreatment group. As in the original MTA study, we also tested baseline, Pearson’s correlations of irritability with ADHD
for site differences and site-by-treatment effects using the interaction symptoms (r ¼ 0.34; p < .001) and with the CISP (r ¼ 0.45;
sitetreatment group. In statistical terms, our hypothesis was that p < .001) were in the medium range.
there would be a significant time-by-treatment group interaction, and
that by decomposing this interaction we would find that the MedMgt
group (as well as the Comb) would be superior to the Beh group. In Aim I: To Establish the Independence of the Irritability
addition, we calculated the pre–post effect size of each treatment Dimension in the MTA Sample
option by using Cohen’s d formula (mean score baseline–mean score The 2-factor model (irritability versus headstrong behaviors)
at 14 months) / pooled SD. 3) Finally, we tested whether baseline
showed a better fit with the data (AIC ¼ 8,640.38; BIC ¼
irritability would moderate treatment outcomes, by running an ITT
8,765.94) compared to the 1-factor model (AIC ¼ 8,751.85;
random-effects regression analysis with the severity of the ADHD
symptoms as the outcome and, again, time (including baseline, 3- BIC ¼ 8,873.08), as confirmed by difference testing (value ¼
month, 9-month, and 14-month assessments), treatment group, and 69.23, df ¼ 1; p  .001).
site as predictors, as well as the interactions irritabilitytime, irri- Also, the within-domain stability was significantly
tabilitytreatment group, timetreatment group, sitetreatment stronger than the across-domain stability: irritability at
group, and irritabilitytimetreatment group. We also tested baseline was a better predictor of irritability at 14 months
whether irritability would differentially affect the response to in- (b ¼ 0.52 [95% CI ¼ 0.42–0.61]; p  .001) compared to
dividual treatments by testing the 3-way interaction of irrita- headstrong behaviors at 14 months (b ¼ 0.15 [95% CI ¼ 0.05–
bilitytreatment grouptime. 0.25]; p ¼ .004), whereas the headstrong behaviors dimen-
sion at baseline was a significantly better predictor of
Ethical Approval headstrong behaviors at 14 months (b ¼ 0.43 [95% CI ¼
The de-identified MTA dataset (MTA96, Version #1) was pro- 0.33.52]; p  .001) than irritability at 14 months (b ¼ 0.05
vided by the National Institute of Mental Health (NIMH) upon [95% CI ¼ 0.06–0.15]; p ¼ .366). This supports the idea that
public-access request. The Psychiatry, Nursing, and Midwifery these constructs are distinct from each other (for more details
Research Ethics Subcommittee (PNM RESC) at King’s College
on the path analytical model, see Figure S1, available online).
London approved the secondary analysis of these data (reference
PNM/13/14-34).
Irritability was a significantly stronger predictor than
headstrong behaviors for the Internalizing Scale at baseline
(irritability: b ¼ 0.43 [95% CI ¼ 0.32–0.54] versus head-
RESULTS strong behaviors: b ¼ 0.06 [95% CI ¼ –0.04–0.17]) as well
The mean score for the whole sample for the parent-reported as at the end of treatment (irritability: b ¼ 0.35 [95%
irritability subscale derived from the ODD items of the CI ¼ 0.23–0.47] versus headstrong behaviors: b ¼ 0.03
SNAP was 4.30 (SD ¼ 2.48) and the median value was 4. [95% CI ¼ –0.15–0.08]). Conversely, the headstrong behav-
Internal consistency was high (Cronbach’s a ¼ .83), and it iors measure was a significantly stronger predictor of the
showed good convergent validity by correlating highly with Externalizing Scale at baseline (irritability: b ¼ 0.29 [95%
a scale comprising irritability items from the CBCL (Pear- CI ¼ 0.21–0.38] versus headstrong behaviors: b ¼ 0.47 [95%
son’s correlation r ¼ 0.66, latent correlation r ¼ 0.72). This CI ¼ 0.38–0.55]) but not at the end of treatment (irritability:
measure of irritability derived from the CBCL has also been b ¼ 0.22 [95% CI ¼ 0.10–0.33] versus headstrong behaviors:
used in previous research.17 The mean score for the whole b ¼ 0.29 [95% CI ¼ 0.18–0.40]) (see Table S1, available online,
sample for the parent-reported headstrong behaviors sub- for more details).
scale derived from the ODD items of the SNAP was 6.44 Finally, irritability and headstrong behaviors each
(SD ¼ 3.19), and the median value was 7. This variable also contributed independently to impairment (irritability:
showed high internal consistency (Cronbach’s a ¼ 0.83). At b ¼ 0.24, p < .001; headstrong behaviors: b ¼ 0.30, p  .001).

TABLE 2 Irritability Response to Attention-Deficit/Hyperactivity Disorder Treatment

Outcome: Parent-Reported Irritability c2 Coefficient CI P
Omnibus Tests
Time 44.52 — — .000
Treatment Group 1.38 — — .711
TimeTreatment Group 33.66 — — .000
Decomposing TimeTreatment Group Interaction
CC vs. MedMgt — 0.42 (0.92 to 0.07) .095
CC vs. Beh — 0.16 (0.33 to 0.65) .515
CC vs. Comb — 0.71 (1.20 to 0.22) .004
MedMgt vs. Beh — 0.58 (0.09 to 1.01) .021
MedMgt vs. Comb — 0.29 (0.78 to 0.20) .249
Beh vs. Comb — 0.87 (0.39 to 1.36) .000
Note: Significant results are shown in boldface. Beh ¼ behavioral treatment; CC ¼ community comparison; Comb ¼ combined treatment; MedMgt ¼ Medication
management.

JOURNAL OF THE AMERICAN ACADEMY OF C HILD & ADOLESCENT PSYCHIATRY

VOLUME 54 NUMBER 1 JANUARY 2015 www.jaacap.org 65
 
FERNANDEZ DE LA CRUZ et al.

Aim II: To Test Whether ADHD Treatments Are Effective FIGURE 2 Changes in parent-reported attention-deficit/
at Treating Irritability hyperactivity disorder (ADHD) scores in the 4 treatment groups
There were no differences in the level of parent-reported in individuals with high (a) and low (b) irritability. Note: The
irritability at baseline in the 4 treatment groups (F ¼ 0.43, categorical outcome was generated using a median split into
df ¼ 3, p ¼ .729). As shown in Figure 1, overall, irritability high and low irritability and is used in this figure for purposes
scores decreased over the course of the treatment. Means of illustration. However, a dimensional irritability variable
and standard deviations at baseline and at 14 months, as is used in the statistical models presented in the text. Beh ¼
well as pre–post effect sizes for each treatment arm, can be behavioral treatment; CC ¼ community comparison; Comb ¼
found in Table 1. The highest effect size corresponded to combined treatment; MedMgt ¼ Medication management.
combined treatment (0.82), followed by medication man-
agement (0.63), community comparison (0.48), and behav-
ioral treatment (0.42).
The random-effects regression model included time,
treatment group, and site as predictors of change in irrita-
bility. Omnibus Wald tests for the effect of site did not reach
significance (site: c2 [5] ¼ 10.61, p ¼ .060; sitetreatment: c2
[15] ¼ 18.84; p ¼ .221); the variable was therefore excluded in
a subsequent, more parsimonious model, which included
time and treatment group only as predictors of change
(Table 2). Effects of time were significant, indicating that
irritability scores decreased significantly over time. Effects of
treatment group did not reach significance, indicating that
irritability scores in the 4 groups were not significantly
different across groups. The interaction time
treatment group was significant, indicating that irritability
scores changed differentially according to treatment groups.
Decomposition of this interaction (Table 2) indicated that
from the baseline to the end of the treatment (14-month
assessment), the MedMgt group were more likely to
improve than Beh (coefficient ¼ 0.58; p ¼ .021), but not the
CC group (coefficient ¼ 0.42; p ¼ .095). Those in the Comb
group were more likely to improve than those in the Beh
(coefficient ¼ 0.87; p ¼ .000) and CC (coefficient ¼ 0.71;
p ¼ .004) groups, but there was no difference between the
MedMgt only and Comb (coefficient ¼ 0.29; p ¼ .249). It is
important to note that the effect sizes of difference between
treatments are more modest compared to the effect sizes
between pre- and posttreatment for each group. For timetreatment, and sitetreatment (Table 3). Omnibus
example, the effect size of the difference between the Comb Wald tests for the effect of site did not reach significance
and CC arms is 0.34, whereas the difference between CC and (site: c2 [5] ¼ 8.29, p ¼ .141; sitetreatment: c2 [15] ¼ 14.15;
Beh is 0.05. However, it is worth noting that the magnitude p ¼ .514), for which the variable was excluded in a subse-
of the difference between ADHD treatments is best quent, more parsimonious model, which included time,
accounted for by the random-effects regression model treatment group, and irritability only as predictors of change
which, unlike the effect sizes, takes into account the het- (Table 3). In addition, we estimated a 3-way interaction
erogeneity across individuals in their responses over time. model including timetreatment groupirritability for
Effect sizes are displayed in Table 1. ADHD symptoms as the outcome. As expected, results of
the 3-way interaction were not significant (coefficients
ranging from 0.02 to 0.01; p values ranging from .481 to
Aim III: To Test Whether Irritability Moderates Treatment .898; full results of this model are presented in Table S2,
Response of Children With ADHD available online). Moreover, the first, more simplified model
Irritability (displayed in Figure 2 as a categorical variable also presented a better fit compared to the model including
using a median split into high and low irritability for illus- the 3-way interaction (AIC ¼ 2924.95 versus 2938.69; BIC ¼
tration but analyzed dimensionally) did not have a differ- 3064.30 versus 3128.21, respectively).
ential effect on the reduction in parent-reported ADHD There was a significant main effect of irritability (reflect-
symptoms across the treatment groups across time. Using ing the higher baseline scores of ADHD symptoms in chil-
the dimensional irritability variable, we formally tested this dren with high baseline irritability) and a main effect of time
in a random-effects regression model that included base- (indicating that ADHD scores decreased significantly over
line irritability, treatment group, time, and site, with time). The effect of site (but not of the interaction
the interactions irritabilitytime, irritabilitytreatment, sitetreatment) was also significant. As shown in Table 3,

JOURNAL OF THE AMERICAN ACADEMY OF C HILD & ADOLESCENT PSYCHIATRY

66 www.jaacap.org VOLUME 54 NUMBER 1 JANUARY 2015
 TREATMENT OF ADHD AND IRRITABILITY

TABLE 3 Attention-Deficit/Hyperactivity Disorder (ADHD) Response to Multimodal Treatments Including Baseline Irritability
as a Factor
Outcome: Parent-Reported ADHD c2 Coefficient CI p
Omnibus Tests
Time 57.90 — — .000
Treatment Group 0.19 — — .979
Irritability 21.06 — — .000
TimeTreatment Group 123.41 — — .000
TimeIrritability 7.50 — — .058
IrritabilityTreatment Group 0.16 — — .984
Decomposing TimeTreatment Group Interaction
CC vs. MedMgt — L0.40 (L0.54 to L0.25) .000
CC vs. Beh — L0.03 (0.17 to 0.11) .681
CC vs. Comb — L0.50 (L0.65 to L0.36) .000
MedMgt vs. Beh — 0.37 (0.22 to 0.51) .000
MedMgt vs. Comb — L0.11 (0.25 to 0.03) .136
Beh vs. Comb — 0.48 (0.33 to 0.62) .000

Outcome: Teacher-Reported ADHD c2 Coefficient CI p

Omnibus Tests
Time 87.06 — — .000
Treatment Group 1.17 — — .620
Irritability 1.05 — — .305
TimeTreatment Group 70.13 — — .000
TimeIrritability 3.10 — — .377
IrritabilityTreatment Group 1.65 — — .648
Decomposing TimeTreatment Group Interaction
CC vs. MedMgt — L0.47 (L0.65 to L0.29) .000
CC vs. Beh — 0.14 (0.32 to 0.04) .125
CC vs. Comb — L0.35 (L0.53 to L0.18) .000
MedMgt vs. Beh — 0.33 (0.15 to 0.51) .000
MedMgt vs. Comb — 0.12 (0.06 to 0.30) .204
Beh vs. Comb — 0.21 (0.03 to 0.39) .022
Note: Significant results are shown in boldface. Beh ¼ behavioral treatment; CC ¼ community comparison; Comb ¼ combined treatment; MedMgt ¼
Medication management.

the interaction treatmenttime reached significance. DISCUSSION

Decomposition of this interaction showed a better response This study used the MTA data to examine irritability in
to treatment for children in the MedMgt and Comb arms children with ADHD and its response to different treat-
compared to those in CC or Beh groups, the same as in the ments. Our findings were in line with our initial hypothesis
original ITT report.2 that, in children with ADHD, irritability is a separable
Additional analyses looked at teacher-rated ADHD as an dimension within the ODD construct. A number of previous
outcome in the previous model. This was to ensure that the studies converge in showing that oppositionality is best
paper’s main findings held across informant sources, but thought of as comprising 2 (irritable and headstrong)17,36 or
also because teachers were more likely to be blinded to the 3 (irritable, headstrong, and hurtful)14-16,37 dimensions with
treatment condition. As in the case of the parent-rated distinct correlates. We did not have enough items to
ADHD, the variable site did not reach significance (site: c2 examine the hurtful dimension in this study, although a
[5] ¼ 5.33, p ¼ .377; sitetreatment: c2 [15] ¼ 12.54; p ¼ .638) previous study suggests that this does exist in children
and therefore was dropped from the model. As can be seen with ADHD.38 Consistent with previous studies, irritabil-
in the lower part of Table 3, results were very similar to ity in this sample contributed to impairment, was more
those using the parent-rated ADHD as an outcome, except associated with emotional than conduct problems, and
for the fact that the main effect of irritability did not reach showed longitudinal continuity. Thus, irritability in ADHD
significance in this case. As in the case of the parent-reported has the same pattern of multivariate structure and corre-
ADHD, results of the 3-way interaction timetreatment lates as in children without ADHD. This lends support to
groupirritability were not significant for the teacher report the notion that, instead of irritability being an ADHD-
(see Table S3, available online). specific phenomenon, it is a dimension that cuts across

JOURNAL OF THE AMERICAN ACADEMY OF C HILD & ADOLESCENT PSYCHIATRY

VOLUME 54 NUMBER 1 JANUARY 2015 www.jaacap.org 67
 
FERNANDEZ DE LA CRUZ et al.

psychopathology in the manner of the Research Domain the results were similar to those obtained when using the
Criteria (RDoC) conceptualization.39 However, further parent reports. Therefore, considering that teachers were
studies will be required to determine whether the etiolog- probably blinded (i.e., unlikely to be aware of the treatment
ical mechanisms underlying irritability differ between in- allocation),23 it is unlikely that parental expectations played
dividuals with and without ADHD. a relevant role in the teacher-rated results, although it is
Our second aim was to test the hypothesis that symptoms possible that parental expectations affected the child’s
of irritability would diminish with medication. We found behavior in a way that carried over into school. Third, it may
that irritability levels decreased in all treatment arms after 14 well be that we were underpowered to detect a 3-way
months. However, the magnitude of the effect sizes for the interaction (irritabilitytimetreatment group) in our
irritability response to treatment was approximately half of models. However, based on previous simulation results, we
the magnitude for ADHD symptoms in the original study.2 have estimated that the sample size required to detect dif-
In support of our hypothesis, MedMgt was significantly ferences among the groups in such a 3-way interaction
better in reducing irritability than Beh treatment. Surpris- would be more than 7,000 participants, which is unrealistic
ingly, MedMgt was not significantly better than CC in for most clinical trials.43,44 On the other hand, it is reassuring
reducing irritability (p ¼ .095). However, combining that the graphs shown in Figure 2 did not suggest the
MedMgt with Beh treatment was superior to both the Beh presence of this interaction, and if the moderating effect is so
treatment alone and the CC intervention, but not compared small, it is unlikely to be clinically meaningful. Finally, it is
to MedMgt alone. Beh treatment was not significantly worth noting that the community care arm in the MTA
different from the community care intervention. These re- study presented with high medication levels (>70% of chil-
sults show a partial overlap with previous MTA findings dren were taking medication for ADHD, albeit with much
regarding other disruptive symptoms. Jensen et al.40 less consistency/monitoring and lower dosing than the
analyzed the response to treatment of oppositional and MTA-medicated children) and therefore comparisons
aggressive behaviors and found that not only the combined against this group should be interpreted with caution.
treatment, but also MedMgt and Beh treatments alone, were These results have 2 important clinical implications. First,
each superior to the CC intervention. Moreover, in that stimulants—a commonly used and relatively safe class of
study there were no differences among the 3 active treat- drugs for ADHD—are also helpful for improving irritability
ments for oppositional/aggressive symptoms except for the in children with ADHD. Moreover, the combination of
fact that the Comb was superior to the Beh intervention.40 stimulants and behavioral treatment could help reduce these
This treatment response difference between irritable and symptoms further. Second, irritability symptoms did not
oppositional/aggressive behaviors further suggests that it is have a negative effect on ADHD treatment outcomes. Cli-
important to distinguish between these domains. In light of nicians can proceed with confidence that ADHD treatments
our findings, it is a possibility that, in the case of irritability, will be effective even in the presence of irritability. These 2
combining medication with behavioral treatment confers aspects had not been demonstrated in major randomized
advantages given the superiority of the combination, but not controlled trials in the field and have long remained an area
MedMgt alone, over the CC, although our results did not of clinical uncertainty. Our results may also have etiological
actually show superiority of the Comb over MedMgt. Also, implications. Based on the fact that irritability improves with
our study results indicate that not all irritability remits after treatments that are effective for ADHD symptoms, it would
standard treatment. Whether adjunct treatments, for be tempting to assume that common pathophysiology un-
example those that have been shown to be effective in derlies the overlap between irritability and ADHD. Further
treating aggression in ADHD,41 should be considered for research should therefore explore this possibility. Also,
irritability remains to be established. future studies should investigate whether ADHD treatment
Finally, we wanted to examine whether high levels of in children with irritable symptoms has a beneficial impact
irritability diminished ADHD treatment response. We found on mood symptoms in the medium-to-long term, given the
that the combined and medication-only treatment arms were links between irritability and mood disorders.45 &
superior to the behavioral treatment and the community
care interventions at reducing ADHD symptoms, regardless
of the level of irritability. This is in keeping with previous
reports on the MTA data showing that the comorbidity of
Clinical Guidance
ADHD with ODD or CD rarely interacted with treatment
response or outcomes.10,42
 Irritability is a separable dimension within the ODD
Limitations of this study include the fact that the MTA
construct in children with ADHD.
was not originally designed to examine irritability in chil-
dren with ADHD, and therefore patient randomization was  Standard ADHD treatments are helpful for reducing
not stratified by irritability status. Second, the parents in the irritability in children with ADHD.
MTA study were not blinded to treatment group assign-  Irritability symptoms do not seem to influence ADHD
ments. As such, the extent to which differential outcomes as treatment outcomes. Clinicians can proceed with
a function of treatment group were influenced by parental confidence that ADHD treatments will be effective even in
expectations cannot be determined. However, teacher re- the presence of irritability.
ports of ADHD symptoms were also used as outcomes, and

JOURNAL OF THE AMERICAN ACADEMY OF C HILD & ADOLESCENT PSYCHIATRY

68 www.jaacap.org VOLUME 54 NUMBER 1 JANUARY 2015
 TREATMENT OF ADHD AND IRRITABILITY

Accepted October 17, 2014. Glen R. Elliott, PhD, MD (San Francisco); Duke University: C. Keith Conners,
Drs. Fern andez de la Cruz, Simonoff, and Stringaris are with the Institute of PhD, Karen C. Wells, PhD, John March, MD, MPH; University of California,
Psychiatry, King’s College London, UK. Dr. McGough is with the Semel Irvine/Los Angeles: James Swanson, PhD (Irvine), Dennis P. Cantwell, MD,
Institute for Neuroscience and Human Behavior at the University of California, (deceased, Los Angeles), Timothy Wigal, PhD (Irvine); Long Island Jewish
Los Angeles, USA. Dr. Halperin is with Queens College, City University of Medical Center/Montreal Children’s Hospital: Howard B. Abikoff, PhD
New York, New York City, USA. Dr. Arnold is with Nisonger Center, The (currently at New York University School of Medicine), Lily Hechtman, MD
Ohio State University, Columbus, OH, USA. (McGill University); New York State Psychiatric Institute/Columbia University/
Mount Sinai Medical Center: Laurence L. Greenhill, MD (Columbia), Jeffrey H.
Dr. Fern andez de la Cruz was fully supported and Dr. Simonoff was partly Newcorn, MD (Mount Sinai School of Medicine); University of Pittsburgh:
supported by the National Institute for Health Research (NIHR) Biomedical William E. Pelham, PhD (currently at Florida International University), Betsy
Research Centre for Mental Health at South London and Maudsley National Hoza, PhD (currently at University of Vermont). Statistical and design consul-
Health Service (NHS) Foundation Trust and the Institute of Psychiatry, King’s tant: Helena C. Kraemer, PhD (Stanford University). Collaborator from the
College London. The Multimodal Treatment Study of Children With ADHD Office of Special Education Programs/US Department of Education: Ellen
(MTA) has been supported by the following grant numbers: U01MH50440, Schiller, PhD.
U01MH50447, U01MH50453, U01MH50454, U01MH50461,
U01MH50467, and the following contract numbers: N01MH12004, Disclosure: Dr. McGough has received grant or research support from NIH,
N01MH12007, N01MH12008, N01MH12009, N01MH12010, NeuroSigma, Inc., Purdue Pharma L.P., and Shire Pharmaceuticals. He has
N01MH12011, N01MH12012; HHSN271200800003-C, HHSN27 served as a consultant to Akili Interactive Labs, Merck, Neurovance, and
1200800004-C, HHSN271200800005-C, HHSN271200800006-C, Sunovion. He has presented expert testimony for Shire. Dr. Arnold has
HHSN271200800007-C, HHSN271200800008-C, and HHSN27 received research funding from Curemark, Forest, Eli Lilly and Co., Neuro-
1200800009-C. This article presents independent research funded by the pharm, Novartis, Noven, Shire, Young Living, NIH, and Autism Speaks, and
NIHR. The views expressed are those of the authors and not necessarily those has consulted with or been on advisory boards for Gowlings, Neuropharm,
of the Department of Health, the MTA Study Investigators, the National In- Novartis, Noven, Organon, Otsuka, Pfizer, Roche, Seaside Therapeutics,
stitutes of Health (NIH), the NHS, or the NIHR. Sigma Tau, Shire, and Tris Pharma, and has received travel support from
John Hodsoll, PhD, of King’s College London served as the statistical expert for Noven. Dr. Stringaris has received grant or research support from the Well-
this research. come Trust, the NIHR, and the Department of Health UK. He has received
The authors would like to thank Pablo Vidal-Ribas, MSc, of King’s College royalties from Cambridge University Press for his book The Maudsley Reader in
London for important insights during analysis of the data. Data used in the Phenomenological Psychiatry. Drs. Fern andez de la Cruz, Simonoff, and
preparation of this article (MTA96, version 1) were obtained from the limited Halperin report no biomedical financial interests or potential conflicts of in-
access datasets of the MTA, which was a National Institute of Mental Health terest.
(NIMH) cooperative agreement randomized clinical trial involving 6 clinical Correspondence to Lorena Fern andez de la Cruz, PhD, Mood and Develop-
sites. Collaborators from NIMH: Peter S. Jensen, MD (currently at REACH ment Lab, Department of Child and Adolescent Psychiatry, King’s College
Institute and Mayo Clinic), L. Eugene Arnold, MD, MEd (currently at Ohio London, Institute of Psychiatry; PO Box 85, De Crespigny Park; London SE5
State University), Joanne B. Severe, MS (Clinical Trials Operations and
8AF, UK; e-mail: [email protected]
Biostatistics Unit, Division of Services and Intervention Research), Benedetto
Vitiello, MD (Child and Adolescent Treatment and Preventive Interventions 0890-8567/$36.00/ª2015 American Academy of Child & Adolescent
Research Branch), John Richters, PhD (currently at National Institute of Nursing Psychaitry. Published by Elsevier Inc. This is an open access article under the
Research), Donald Vereen, MD (currently at National Institute on Drug Abuse). CC BY license (http://creativecommons.org/licenses/by/3.0/).
Principal investigators and co-investigators from the 6 sites were: University of
California, Berkeley/San Francisco: Stephen P. Hinshaw, PhD (Berkeley), http://dx.doi.org/10.1016/j.jaac.2014.10.006

REFERENCES
1. Shaw P, Stringaris A, Nigg J, Leibenluft E. Emotion dysregulation in treatment: impact on parental practices. J Am Acad Child Adolesc Psy-
attention deficit hyperactivity disorder. Am J Psychiatry. 2014;171: chiatry. 2004;43:830-838.
276-293. 12. Daley D, Van der Oord S, Ferrin M, et al. Behavioral interventions in
2. MTA Cooperative Group. A 14-month randomized clinical trial of attention-deficit/hyperactivity disorder: a meta-analysis of randomized
treatment strategies for attention-deficit/hyperactivity disorder. Arch controlled trials across multiple outcome domains. J Am Acad Child
Gen Psychiatry. 1999;56:1073-1086. Adolesc Psychiatry. 2014;53:835-847.
3. Polanczyk G, de Lima MS, Horta BL, Biederman J, Rohde LA. The 13. Aebi M, Muller UC, Asherson P, et al. Predictability of oppositional
worldwide prevalence of ADHD: a systematic review and metare- defiant disorder and symptom dimensions in children and adolescents
gression analysis. Am J Psychiatry. 2007;164:942-948. with ADHD combined type. Psychol Med. 2010;40:2089-2100.
4. American Psychiatric Association. Diagnostic and Statistical Manual of 14. Stringaris A, Goodman R. Longitudinal outcome of youth opposition-
Mental Disorders. 5th edition (DSM-5). Washington, DC: American ality: irritable, headstrong, and hurtful behaviors have distinctive pre-
Psychiatric Association; 2013. dictions. J Am Acad Child Adolesc Psychiatry. 2009;48:404-412.
5. Leibenluft E. Severe mood dysregulation, irritability, and the diagnostic 15. Whelan YM, Stringaris A, Maughan B, Barker ED. Developmental con-
boundaries of bipolar disorder in youths. Am J Psychiatry. 2011;168: tinuity of oppositional defiant disorder subdimensions at ages 8, 10, and
129-142. 13 years and their distinct psychiatric outcomes at age 16 years. J Am
6. Still GF. Some abnormal psychical conditions in children: excerpts from Acad Child Adolesc Psychiatry. 2013;52:961-969.
three lectures. J Atten Disord. 2006;10:126-136. 16. Stringaris A, Goodman R. Three dimensions of oppositionality in youth.
7. Barkley RA. Behavioral inhibition, sustained attention, and executive J Child Psychol Psychiatry. 2009;50:216-223.
functions: constructing a unifying theory of ADHD. Psychol Bull. 1997; 17. Stringaris A, Zavos H, Leibenluft E, Maughan B, Eley TC. Adolescent
121:65-94. irritability: phenotypic associations and genetic links with depressed
8. Stringaris A, Goodman R. Mood lability and psychopathology in youth. mood. Am J Psychiatry. 2012;169:47-54.
Psychol Med. 2009;39:1237-1245. 18. Ahmann PA, Waltonen SJ, Olson KA, Theye FW, Van Erem AJ,
9. Greene RW, Biederman J, Zerwas S, Monuteaux MC, Goring JC, LaPlant RJ. Placebo-controlled evaluation of Ritalin side effects. Pediat-
Faraone SV. Psychiatric comorbidity, family dysfunction, and social rics. 1993;91:1101-1106.
impairment in referred youth with oppositional defiant disorder. Am J 19. Childress AC, Arnold V, Adeyi B, et al. The effects of lisdexamfetamine
Psychiatry. 2002;159:1214-1224. dimesylate on emotional lability in children 6 to 12 years of age with
10. Jensen PS, Hinshaw SP, Kraemer HC, et al. ADHD comorbidity findings ADHD in a double-blind placebo-controlled trial. J Atten Disord. 2014;18:
from the MTA study: comparing comorbid subgroups. J Am Acad Child 123-132.
Adolesc Psychiatry. 2001;40:147-158. 20. Coghill D. Adding multimodal behavioural therapy to methylphenidate
11. Hechtman L, Abikoff H, Klein RG, et al. Children with ADHD treated does not improve ADHD outcomes. Evid Based Ment Health. 2007;
with long-term methylphenidate and multimodal psychosocial 10:124.

JOURNAL OF THE AMERICAN ACADEMY OF C HILD & ADOLESCENT PSYCHIATRY

VOLUME 54 NUMBER 1 JANUARY 2015 www.jaacap.org 69
 
FERNANDEZ DE LA CRUZ et al.

21. Manos MJ, Brams M, Childress AC, Findling RL, Lopez FA, Jensen PS. 34. Bird HR, Shaffer D, Fisher P, et al. The Columbia Impairment Scale
Changes in emotions related to medication used to treat ADHD. Part I: (CIS)—pilot findings on a measure of global impairment for children and
literature review. J Atten Disord. 2011;15:101-112. adolescents. Int J Methods Psychiatr Res. 1993;3:167-176.
22. Mongia M, Hechtman L. Cognitive behavior therapy for adults with 35. Stringaris A, Goodman R. The value of measuring impact along-
attention-deficit/hyperactivity disorder: a review of recent randomized side symptoms in children in adolescents: a longitudinal assess-
controlled trials. Curr Psychiatry Rep. 2012;14:561-567. ment in a community sample. J Abnorm Child Psychol. 2013;41:
23. Sonuga-Barke EJ, Brandeis D, Cortese S, et al. Nonpharmacological in- 1109-1120.
terventions for ADHD: systematic review and meta-analyses of ran- 36. Rowe R, Costello EJ, Angold A, Copeland WE, Maughan B. Develop-
domized controlled trials of dietary and psychological treatments. Am J mental pathways in oppositional defiant disorder and conduct disorder.
Psychiatry. 2013;170:275-289. J Abnorm Psychol. 2010;119:726-738.
24. Galanter CA, Carlson GA, Jensen PS, et al. Response to methylphenidate 37. Krieger FV, Polanczyk VG, Robert G, et al. Dimensions of opposition-
in children with attention deficit hyperactivity disorder and manic ality in a Brazilian community sample: testing the DSM-5 proposal
symptoms in the Multimodal Treatment Study of Children with Atten- and etiological links. J Am Acad Child Adolesc Psychiatry. 2013;52:
tion Deficit Hyperactivity Disorder titration trial. J Child Adolesc Psy- 389-400. e381.
chopharmacol. 2003;13:123-136. 38. Kolko DJ, Pardini DA. ODD dimensions, ADHD, and callous-
25. Galanter CA, Pagar DL, Davis M, et al. ADHD and manic symptoms: unemotional traits as predictors of treatment response in children with
Diagnostic and treatment implications. Clinical Neuroscience Research. disruptive behavior disorders. J Abnorm Psychol. 2010;119:713-725.
2005;5:283-294. 39. Insel T, Cuthbert B, Garvey M, et al. Research domain criteria (RDoC):
26. Arnold LE, Abikoff HB, Cantwell DP, et al. NIMH collaborative multi- toward a new classification framework for research on mental disorders.
modal treatment study of children with ADHD (MTA): design, meth- Am J Psychiatry. 2010;167:748-751.
odology, and protocol evolution. J Atten Disord. 1997;2:141-158. 40. Jensen PS, Hinshaw SP, Swanson JM, et al. Findings from the NIMH
27. Arnold LE, Abikoff HB, Cantwell DP, et al. National Institute of Mental Multimodal Treatment Study of ADHD (MTA): implications and
Health Collaborative Multimodal Treatment Study of Children with applications for primary care providers. J Dev Behav Pediatr. 2001;
ADHD (the MTA). Design challenges and choices. Arch Gen Psychiatry. 22:60-73.
1997;54:865-870. 41. Blader JC, Schooler NR, Jensen PS, et al. Adjunctive divalproex versus
28. Hinshaw S, March JS, Abikoff H, et al. Comprehensive assessment of placebo for children with ADHD and aggression refractory to stimulant
childhood attention-deficit hyperactivity disorder in the context of a monotherapy. Am J Psychiatry. 2009;166:1392-1401.
multisite, multimodal clinical trial. J Atten Disord. 1997;1:217-234. 42. The MTA Cooperative Group. Moderators and mediators of treatment
29. Swanson JM. School-Based Assessments and Interventions for ADD response for children with attention-deficit/hyperactivity disorder: the
Students. Irvine, CA: KC Publications; 1992. Multimodal Treatment Study of Children with Attention Deficit Hyper-
30. Stringaris A, Goodman R. Mood lability and psychopathology in youth. activity Disorder. Arch Gen Psychiatry. 1999;56:1088-1096.
Psychol Med. 2009;39:1237-1245. 43. Landau S, Stahl D. Sample size and power calculations for medical
31. Loeber R, Green SM, Lahey BB. Mental-health professionals perception of studies by simulation when closed form expressions are not available.
the utility of children, mothers, and teachers as informants on childhood Stat Methods Med Res. 2013;22:324-345.
psychopathology. J Clin Child Psychol. 1990;19:136-143. 44. Lenth RV. Post Hoc Power: Tables and Commentary. Iowa City:
32. Phares V. Accuracy of informants: do parents think that mother knows Department of Statistics and Actuarial Science, University of
best? J Abnorm Child Psychol. 1997;25:165-171. Iowa; 2007.
33. Achenbach TM. Manual for the Child Behavior Checklist/4-18 and 1991 45. Maughan B, Collishaw S, Stringaris A. Depression in childhood
Profile. Department of Psychiatry. Burlington, VT: University of Ver- and adolescence. J Can Acad Child Adolesc Psychiatry. 2013;
mont; 1991. 22:35-40.

JOURNAL OF THE AMERICAN ACADEMY OF C HILD & ADOLESCENT PSYCHIATRY

70 www.jaacap.org VOLUME 54 NUMBER 1 JANUARY 2015
VOLUME 54 NUMBER 1 JANUARY 2015
JOURNAL OF THE AMERICAN ACADEMY OF C HILD & ADOLESCENT PSYCHIATRY

TABLE S1 Linear Regressions With the Irritability and Headstrong Dimensions as Predictors and the Child Behavior Checklist (CBCL) Subscales as Outcomes
Predictors

Outcome Irritability Dimension Headstrong Dimension

CBCL Difference Between

Subscales b 95% CI p b 95% CI p Coefficient 95% CI Predictors (p)
Withdrawn Baseline 0.32 (0.21 to 0.43) .009 0.04 (0.07 to 0.16) .438 L0.27 (L0.48 to 0.07) .009
14 mo 0.25 (0.13 to 0.37) .027 0.00 (0.12 to 0.12) .958 L0.25 (L0.48 to 0.03) .027
Somatic Complaints Baseline 0.16 (0.04 to 0.27) .678 0.11 (0.00 to 0.23) .059 0.04 (0.26 to 0.17) .678
14 mo 0.18 (0.06 to 0.31) .087 0.01 (0.14 to 0.11) .807 0.20 (0.43 to 0.03) .087
Anxious/Depressed Baseline 0.49 (0.38 to 0.59) .000 0.01 (0.09 to 0.11) .841 L0.48 (L0.66 to L0.28) .000
14 mo 0.38 (0.26 to 0.50) .000 0.05 (0.17 to 0.07) .405 L0.43 (L0.65 to 0.21) .000
Social Problems Baseline 0.17 (0.06 to 0.28) .652 0.22 (0.11 to 0.33) .000 0.05 (0.15 to 0.25) .652
14 mo 0.12 (0.00 to 0.24) .604 0.06 (0.06 to 0.18) .348 0.06 (0.29 to 0.17) .604
Thought Problems Baseline 0.18 (0.07 to 0.30) .484 0.11 (0.01 to 0.22) .071 0.08 (0.29 to 0.14) .484
14 mo 0.10 (0.02 to 0.22) .718 0.06 (0.06 to 0.18) .335 0.04 (0.27 to 0.18) .718
Attention Problems Baseline 0.19 (0.07 to 0.30) .603 0.13 (0.02 to 0.24) .024 0.05 (0.26 to 0.15) .603
14 mo 0.15 (0.03 to 0.28) .202 0.01 (0.12 to 0.13) .912 0.15 (0.38 to 0.08) .202
Delinquent Behavior Baseline 0.10 (0.00 to 0.20) .000 0.44 (0.34 to 0.54) .000 0.34 (0.16 to 0.53) .000
14 mo 0.10 (0.01 to 0.22) .100 0.28 (0.17 to 0.40) .000 0.18 (0.03 to 0.39) .100
Aggressive Behavior Baseline 0.34 (0.25 to 0.42) .274 0.42 (0.34 to 0.51) .000 0.08 (0.07 to 0.24) .274
14 mo 0.24 (0.12 to 0.35) .738 0.27 (0.16 to 0.38) .000 0.03 (0.17 to 0.24) .738
Sex Problems Baseline 0.06 (0.06 to 0.30) .289 0.18 (0.06 to 0.30) .003 0.12 (0.10 to 0.33) .289
14 mo 0.03 (0.09 to 0.15) .617 0.09 (0.03 to 0.21) .139 0.06 (0.17 to 0.27) .617
Internalizing Scalea Baseline 0.43 (0.32 to 0.54) .000 0.06 (0.04 to 0.17) .244 0.37 (L0.56 to L0.17) .000
14 mo 0.35 (0.23 to 0.47) .001 0.03 (0.15 to 0.08) .575 L0.39 (L0.61 to L0.17) .001

TREATMENT OF ADHD AND IRRITABILITY

Externalizing Scalea Baseline 0.29 (0.21 to 0.38) .026 0.47 (0.38 to 0.55) .000 0.17 (0.02 to 0.32) .026
14 mo 0.22 (0.10 to 0.33) .456 0.29 (0.18 to 0.40) .000 0.07 (0.13 to 0.28) .456
Total Score Baseline 0.35 (0.26 to 0.44) .802 0.33 (0.23 to 0.42) .000 0.02 (0.19 to 0.15) .802
14 mo 0.28 (0.16 to 0.39) .199 0.14 (0.02 to 0.25) .020 0.14 (0.35 to 0.07) .199
www.jaacap.org

Note: Significant results are shown in boldface.

a
The Internalizing Scale is composed of the subscales Withdrawn, Somatic Complaints, and Anxious/Depressed Behavior. The Externalizing Scale is composed of the Delinquent Behavior and Aggressive Behavior
scales.
70.e1
 
FERNANDEZ DE LA CRUZ et al.

TABLE S2 Parent-Reported Attention-Deficit/Hyperactivity Disorder (ADHD) Response to Multimodal Treatments Including Baseline
Irritability as a Factor (3-Way Interaction Model: IrritabilityTimeTreatment Group)
Outcome: Parent-Reported ADHD c2 Coefficient CI p
Omnibus Tests
Time 30.54 — — .000
Treatment Group 0.43 — — .934
Irritability 11.53 — — .001
TimeTreatment Group 20.51 — — .015
TimeIrritability 1.45 — — .695
IrritabilityTreatment Group 0.90 — — .825
TimeIrritabilityTreatment Group 4.27 — — .893
Decomposing TimeTreatment Group Interaction
CC vs. MedMgt — L0.38 (L0.67 to L0.09) .011
CC vs. Beh — 0.01 (0.28 to 0.29) .969
CC vs. Comb — L0.41 (L0.71 to L0.12) .006
MedMgt vs. Beh — 0.43 (0.14 to 0.74) .004
MedMgt vs. Comb — 0.06 (0.25 to 0.37) .702
Beh vs. Comb — 0.38 (0.07 to 0.69) .016
Decomposing IrritabilityTimeTreatment Group
Interaction
CC vs. MedMgt — 0.00 (0.06 to 0.05) .877
CC vs. Beh — 0.01 (0.07 to 0.05) .784
CC vs. Comb — 0.02 (0.08 to 0.04) .481
MedMgt vs. Beh — 0.00 (0.06 to 0.05) .898
MedMgt vs. Comb — 0.02 (0.07 to 0.04) .567
Beh vs. Comb — 0.01 (0.05 to 0.07) .663
Note: Significant results are shown in boldface. Beh ¼ behavioral treatment; CC ¼ community comparison; Comb ¼ combined treatment; MedMgt ¼ medication
management.

JOURNAL OF THE AMERICAN ACADEMY OF C HILD & ADOLESCENT PSYCHIATRY

70.e2 www.jaacap.org VOLUME 54 NUMBER 1 JANUARY 2015
 TREATMENT OF ADHD AND IRRITABILITY

TABLE S3 Teacher-Reported Attention-Deficit/Hyperactivity Disorder (ADHD) Response to Multimodal Treatments Including

Baseline Irritability as a Factor (3-Way Interaction Model: IrritabilityTimeTreatment Group)
Outcome: Teacher-Reported ADHD c2 Coefficient CI p
Omnibus Tests
Time 39.91 — — .000
Treatment Group 2.38 — — .497
Irritability 1.12 — — .291
TimeTreatment Group 15.84 — — .070
TimeIrritability 2.11 — — .551
IrritabilityTreatment Group 1.19 — — .754
TimeIrritabilityTreatment Group 6.51 — — .688
Decomposing TimeTreatment Group Interaction
CC vs. MedMgt — L0.55 (L0.91 to L0.18) .003
CC vs. Beh — 0.24 (0.61 to 0.13) .198
CC vs. Comb — 0.30 (0.68 to 0.07) .111
MedMgt vs. Beh — 0.30 (0.05 to 0.66) .092
MedMgt vs. Comb — 0.24 (0.12 to 0.60) .185
Beh vs. Comb — 0.06 (0.05 to 0.66) .739
Decomposing IrritabilityTimeTreatment Group
Interaction
CC vs. MedMgt — 0.02 (0.05 to 0.09) .644
CC vs. Beh — 0.02 (0.05 to 0.10) .535
CC vs. Comb — 0.01 (0.09 to 0.06) .742
MedMgt vs. Beh — 0.01 (0.07 to 0.08) .853
MedMgt vs. Comb — 0.03 (0.10 to 0.04) .424
Beh vs. Comb — 0.04 (0.04 to 0.11) .346
Note: Significant results are shown in boldface. Beh ¼ behavioral treatment; CC ¼ community comparison; Comb ¼ combined treatment; MedMgt ¼ medication
management.

FIGURE S1 Path analyses of the relation between irritability and headstrong dimensions across time.

JOURNAL OF THE AMERICAN ACADEMY OF C HILD & ADOLESCENT PSYCHIATRY

VOLUME 54 NUMBER 1 JANUARY 2015 www.jaacap.org 70.e3